Your VoIP game plan is key even in data services procurements

Some enterprise network managers truly aren't interested in voice and data convergence. They might not admit it at trade show cocktail parties, but they aren't. Then again, these are the people who probably don't go to trade shows.

This position has merit. Circuit-switched voice works. Toll price-downs have been available year after year. Cost bubbles on mildly value-added services, such as reservationless teleconferencing, have been popped (only a few telecom buyers haven't gotten the word that the premium for this service is very small). Moves/adds/changes on standard telephony are manageable in many companies, despite the advantage that IP telephony provides in this area. Extended apps like call center agents in individual remote locations are cool, but not worth the risk in certain conservative corporate environments, and for reasons other than the technology.

Many telecom professionals are also very wrapped up in issues surrounding the coverage their executives will get over mobile networks nationally and globally. They don't have the time right now to worry about call (or video) quality over fixed-line voice-over-IP. We ourselves at TC2 never promote VoIP for its own sake to clients, and we encourage people to deal with rigorously tested offerings.

So, in 2009, when it comes to procuring network services, can you ignore VoIP, integrated networks and convergence? The answer is: No, you can't.

The most-cited example of this is the PBX market. IP PBXs are rapidly supplanting traditional PBXs in the new-equipment market. Simple examination of the market research indicates you're putting yourself out to pasture if you invest in a purely traditional voice switch, inviting near-term obsolescene in parts and support. Believe me, the market researcher par excellence in this field, veteran consultant Allan Sulkin, is as skeptical of unproven trends as they come, and in the past has warned users away from half-baked technologies that were destined to flounder. But Al's comprehensive annual voice CPE survey for the folks at VoiceCon and NoJitter.com now clearly shows what you need to be thinking about in a new voice equipment buy, and it's IP by a mile.

Still, for many, PBX buys are a once-a-decade (or less) event. The same phenomenon manifests itself in a somewhat different way in the transport network services arena, and you may have to understand it on a more pressing basis than for your voice equipment.

On the surface, IP/MPLS services are "data headliners." You buy them because you want today's generation of WAN connectivity -- or are essentially forced to it by your supplier's policies or pricing signals. No one today makes you run voice over these networks, and most MPLS procurements include a request to the same vendors to bid standard toll voice elements as part of a full competitive telecom refresh.

But unlike past generations of data services that formed some kind of "virtually private" carrier-cloud alternative to private-line networks, the very structure of MPLS rate elements does force you to make a specific determination on your current and projected packet-voice practices. You can't just wave the issue away.

AT&T's current key unmanaged MPLS service, called AVPN, and Verizon's, called Private IP or simply "PIP," have slightly different ways of handling this. The very fact that AVPN and PIP have non-parallel mechanisms actually adds to the pressure to project your voice or other delay-sensitive apps into the intermediate term. They may be different mechanisms, but they're both "Class of Service" rate elements -- the very name of which implies a concern for packet-based application performance ranging from "best efforts" Internet browsing to jitter-intolerant voice. That's as opposed to frame relay's "Committed Information Rate" measurement of simply a raw bandwidth reservation across the cloud against competing users.

In a nutshell, AVPN offers a range of CoS packages defined as high or low levels of "multimedia" or "critical data" oriented priorities, and you have to pick pre-determined "profiles" that allocate the port bandwidth by percentage somewhere in this range. PIP's mechanism appears on first blush to be a more binary selection -- you specify whether or not you want a "gold" level of CoS service (called "Gold CARs" after the legacy MCI's use of the term "Committed Access Rate" in early MPLS services).

But in fact, even the choice of a Gold CAR for a particular PIP network site then requires you to specify the precise amount of bandwidth you want for this real-time CoS, which could be 10%, or 20%, or 40% of the full port, representing your expectation of the proportion of traffic that will require this real-time/voice throughput. Significantly, you can't be inconsistent in the way that you present this to AT&T, Verizon and others, and in comparing bids you need to understand where the service definitions of the two services match up. In fact, this may be the biggest challenge in MPLS deal analysis, and you may find it difficult to confidently choose an MPLS service unless you're sure you've poured your application requirements correctly into each proposal.

I know there's a lot of acronyms to deal with here. But the bottom line is that from now on, carrier network services are going to be defined directly around the issue of application performance, in a world where "application" inherently means the entire range of voice, video and data transmissions. Whether you actually run any given application over the network is your choice. But when you buy the service, its potential to run all those applications is key to understanding the service pricing across the range of your bidders' offerings.

The more you are prepared internally with your VoIP roadmap, the better off you'll be in your "data" procurements -- because it's all a network to your suppliers, and the industry as a whole.